PRECIPITATION MEASUREMENT
3.11 OBSERVATIONS OF RAINFALL BY SATELLITE
3.11.1 Basics
Rainfall estimation from space is based on meas-uring the amount of radiation that is reflected and emitted through cloud tops. Most of the radiation does not penetrate deep into cloud regions containing particles with similar or greater size than the radiation wavelength.
Therefore, except for the longest wavelengths, most of the radiation comes from the upper regions of precipitating clouds and can therefore only indirectly be related to surface rainfall.
Consequently there are very many techniques using a range of procedures.
3.11.2 Visible and infra-red
Rain intensities vary with the rate of expansion of cold (T < 235°K) cloud top areas. It is assumed that the expansion of the cloud top is an indicator for the divergence aloft and, hence, to the rate of rising
air and precipitation. However, when used over a large area, this method does not show signifi cant improvement with respect to the simplest possible method which assumes that all clouds with tops colder than a given threshold temperature T precip-itate at a fi xed rate G mm h–1, where T = 235°K and G = 3 mm h–1 is typical of the eastern equatorial Atlantic. This method was developed into the Global Precipitation Index (GPI), which has been used extensively.
Such area methods work well only for a time-space domain that is large enough to include a large number of storms that provide a good representa-tion of the full evolurepresenta-tion of convective rain-cloud systems (for example, 2.5° x 2.5° x 12 hours).
Classifi cation of clouds into convective and strati-form by the texture of the cloud top temperature has shown some improvement for tropical rainfall over land. However, it fails (along with the rest of the infra-red methods) in mid-latitude winter systems, because there the “convective” relation between cold cloud top area and surface rainfall does not apply to largely non-convective cloud systems.
The use of visible wavelengths to characterize the strength of convection works well when used with infra-red wavelengths to indicate the height of the cloud. However, such procedures can be misleading by the presence of bright cirrus cloud, or the pres-ence of low-level orographic rain.
The atmospheric window of about 10 microns is split into two closely spaced wavebands, centred at 10.8 and 12 microns. Clouds have large absorption and emissivity in the longer waveband. Therefore, the 10.8 micron radiation in thin clouds will be contributed from lower and warmer levels as compared with the 12 micron waveband, creating a brightness temperature difference between the two channels. It has been shown that cirrus clouds can be distinguished from thicker cloud by having larger brightness temperature difference. This helps in eliminating thin clouds from consideration as precipitating cloud.
Very cold cloud top temperature is not always a requirement for precipitation, in which case the infra-red threshold technique breaks down. The precipitation formation processes require the exist-ence of large cloud droplets and/or ice particles in the cloud, which often spread to the cloud top.
These large particles absorb the 1.6 and 3.7 micron radiation much more strongly than small cloud droplets. This effect makes it possible to calculate the effective radius (reff = integral volume divided Figure I.3.11. Time series of path-integrated
rainfall estimates from raingauges (solid line) and dual frequency microwave link (dashed line)
attenuation measurements over north-west England on 10 February 2000
(Holt and others, 2000)
Time (hours)
______ Raingauges - - - Microwave link Rainfall rate (mm hr–1)
4 5 6 7 8 9 10
0 2 4 8 10 16 18 20
6 12 14
by integral surface area) of the particles. It has been shown that reff = 14 microns can serve to delineate precipitating clouds, regardless of their top temper-ature (Figure I.3.12).
3.11.3 Passive microwave
Microwaves provide the measurements that are physically best related to the actual precipitation, especially in the longest wavebands. The interac-tions of passive microwave with precipitation clouds and the surface are illustrated in Figure I.3.13, using two wavebands, shorter (85 GHz) and longer (19 GHz). Measurement tech-niques are based on the two physical principles of absorption and scattering.
Absorption-based measurements
Water drops have relatively large absorption/
emission coefficient, increasing for the higher frequencies. The emission is proportional to the vertically integrated cloud and rainwater in the low frequencies, but due to the increased emissivity for the higher frequencies the emission saturates for light rain intensities.
Scattering-based measurements
Ice particles have relatively small absorption/emis-sion, but they are good scatterers of the microwave radiation, especially at higher frequencies.
Therefore, at high frequencies (85 GHz), the large scattering from the ice in the upper portions of the clouds makes the ice an effective insulator, because it refl ects back down most of the radiation emitted from the surface and from the rain. The remaining radiation that reaches the microwave sensor is interpreted as a colder brightness temperature. A major source of uncertainty for the scattering-based retrievals is the lack of a consistent relationship between the frozen hydrometeors aloft and the rainfall reaching the surface.
The two physical principles of absorption and scat-tering described above have been used to formulate a large number of rain estimation methods. In general, passive microwave rainfall estimates over the ocean were of useful accuracy. However, over the equatorial Pacifi c, passive microwave does not show signifi cantly improved skill when compared with the simplest infra-red method (GPI).
Over land the passive microwave algorithms can detect rain mainly by the ice scattering mechanism, and this indirect rain estimation method is less accurate. Moreover, rainfall over land from clouds which do not contain signifi cant amounts of ice aloft, goes mostly undetected.
3.11.4 Active microwave (rain radar;
Tropical Rainfall Measurement Mission)
A major limiting factor in the accuracy of passive microwave methods is the large footprint, which causes partial beam fi lling, especially at the higher frequencies. The resolution is greatly improved with the Tropical Rainfall Measurement Mission (TRMM) satellite, with a corresponding improve-ment in the expected accuracy of the microwave rain estimates. The TRMM satellite has a radar trans-mitting at a wavelength of 2.2 cm (active microwave) and microwave radiometers (19 to 90 GHz) (Figure I.3.14). The resolutions of these instruments range from about 1 km for the visible and infra-red radiometer, about 10 km for the microwave radi-ometers and 250 m for the radar. The radar has provided an improvement in the accuracy of instantaneous rain estimates over those previously achieved from space. Since TRMM samples each area between 35 degrees north and south, at best, twice daily, the sampling error is the dominant source of inaccuracy.
0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1
Arad
CRrad< CRmax
Asat
y = 0.03 + 0.6x R = 0.94 T > 245
T < 245 multilayers
Figure I.3.12. Fraction of precipitating area, defi ned by the area with reff ≥ 14 microns (Asat), as
a function of the fraction of precipitating area detected by radar (Arad) for convective clouds.
Windows with multi-layered clouds are marked with crosses, windows with cloud-top
tempera-tures higher than 245 K are marked with solid circles, and windows with cloud-top temperatures
lower than 245 K are marked with hollow circles.
CRsat is the cloud radius parameter and CRmax is the maximum cloud radius for a given depth.
(Rosenfeld and Gutman, 1994; Lensky and Rosenfeld, 1997)
Rain and land surface Rain-cloud
temperature Sea-surface
temperature
Land surface Sea surface
4 8 7
3
1 2
12 10
0˚C
11
1
8
2 5
6
Figure I.3.13. The interaction of high (for example 85 GHz) and low (for example 19 GHz) frequency passive microwave with precipitation clouds and the surface. The width of the vertical columns
repre-sents the intensity of temperature of the upwelling radiation. The illustrated features and their demarcations are: (a) the small emissivity of sea surface for both low (1) and high (2) frequencies;
(b) the large emissivity of land surface for both low (3) and high (4) frequencies: (c) the emission from cloud and rain drops, which increases with vertically integrated liquid water for the low frequency (5), but saturates quickly for the high frequency (6); (d) the signal of the water emissivity at the low frequency is masked by the land surface emissivity (7); (e) the saturated high frequency emission from the rain (8) is not distinctly different from the land surface background (4); (f) ice precipitation particles
aloft backscatter down the high-frequency emission (9), causing cold brightness temperatures (10), regardless of surface emission properties; (g) the ice lets the low frequency emission upwell unimpeded (11), allowing its detection above cloud top as warm brightness temperature (12).
(Rosenfeld and Collier, 1999)
0 100 200km
.4
0 10 20 30 40 50mm/hr
.8 1.2 1.6 2.0 Inches/hr
Figure I.3.14. Heavy rainfall over Texas derived from the TRMM Microwave Imager and Precipitation Radar on the TRMM satellite at 0439 UTC 1 May 2004 (Courtesy NASA)
A combination of the measurements from TRMM-like and geostationary satellites provides the best potential for accurate global precipitation esti-mates from space. Currently plans are being developed to implement such systems under the general title of the Global Precipitation Mission (GPM).
3.11.5 Summary of accuracy
considerations
In tropical regions there can be a signifi cant diurnal cycle in rainfall activity, and the phase and inten-sity of the cycle may vary from region to region.
The low inclination orbit used for TRMM will proc-ess in such a way as to sample a full diurnal range of Equator crossing times over the course of a month.
This is not the case for satellites in polar orbit for which the Equator crossing time is always the same.
The diurnal cycle may therefore increase the errors due to sampling.
For monthly averages over a 280 km2 and a sampling interval of 10 hours, appropriate for the TRMM satellite, the sampling error is about 10 per cent.
However, for convective systems in other regions, which have shorter decorrelation times than observed for tropical rain, the sampling error is likely to be larger.
The validation of satellite algorithms for estimating rainfall accumulations is complex and must be undertaken in ways that ensure that different tech-niques provide data with similar characteristics, that is, integration times and coverage.
The best accuracy for areal rainfall measurements from space at present is obtained over the tropical oceans, where GPI performs as well as passive microwave techniques for long period (in the order of several months) integrated rainfall. However, errors for individual events may be large because
“warm rain” from shallow clouds is common in some places in the tropics. The passive microwave techniques become increasingly advantageous towards higher latitudes where convective rainfall occurs less frequently. Here the best accuracy is achieved by combining passive microwave with infra-red from geostationary satellites. Somewhat lower accuracies of infra-red techniques are achievable in convective rain over land, due to large dynamic and microphysical diversity of rain-cloud systems. This causes a larger variability between the rainfall and the properties of the upper portions of the clouds. The skill of passive microwave techniques is also reduced over land, because its emissivity reduces greatly the usefulness of frequencies lower
than 35 GHz. Nevertheless, results over land at 88.5 GHz are encouraging.
3.12 REMOTE-SENSING MEASUREMENTS